An Occlusion Detection Algorithm Based on Data Fusion
Occlusion between objects is common in real world and it brings big trouble in environment understanding when a robot movs in urban area. We propose an occlusion detection algorithm which uses both lidar data and image. The method acquires colorized lidar points first by registering ranger finder and camera. Then a fuzzy cluster algorithm based on maximum entropy is employed to segment data points. Each segment is determined whether it is linear distributed or corner distributed by using a multi-time weight least-square fitting algorithm. Occlusion detection is done at last using both geometry and RGB features and segments belong to one object which is assigned the same label. Experiments show the algorithm works well to detect occlusion when objects have regular shape.
occlusion detection data fusion weight least-square fitting robotics
Tian-quan Ni Quan Zhao Xia Yuan
Nanjing University of Aeronautics and Astronautics, Nanjing, China Electric and Electronic Equipment Electric and Electronic Equipment 723 Institute, CSIC, Yangzhou, China Nanjing University of Science and Technology, Nanjing, China
国际会议
The 10th International Conference on Intelligent Technologies(第十届智慧科技国际会议 InTech09)
桂林
英文
608-612
2009-12-12（万方平台首次上网日期，不代表论文的发表时间）